Evaluating the Logical Reasoning Ability of ChatGPT and GPT-4